The role of AI tools in the education of statistical analysis in linguistics: A case study of ChatGPT

This study examines how a large language model (ChatGPT) can serve as a statistical tool in experimental linguistics, compared with conventional tools such as R. Using acceptability judgment data from four student projects with Korean learners of English, we analyze topicalization, wh-dependency, and related structures. In each project, students first obtained ChatGPT-based analyses, and the authors then reanalyzed the same data in R. While ChatGPT offers clear advantages in speed, accessibility, and lowering the entry barrier for learners, we also identify recurring issues in model selection, random-effects specification, multiple-comparison correction, and reproducibility, especially for mixed-effects models. We propose a complementary workflow in which ChatGPT supports early exploration and explanation, while final inferential analysis and reporting are verified in R, highlighting both the pedagogical potential and methodological risks of AI-based statistical analysis in experimental linguistics.